Credit teams have always had access to ratings data. What they have lacked is a way to ask an AI model for it directly and trust the answer came from the rated record, not a guess.
S&P Global said on September 28 that it has integrated its RatingsXpress credit data, covering nearly one million issuer and security-level ratings, outlooks and CreditWatch actions, into the S&P Global AI Data Portal. The portal now supports two access modes: a deterministic, API-driven query for structured workflows that is live now, and an adaptive natural-language mode for more complex, multi-dataset questions arriving in the coming weeks. Analysts can query the ratings data through large language models including GPT, Gemini and Claude rather than pulling reports manually into a separate risk system.
“Credit teams are under pressure to assess risk faster and monitor growing portfolios across sectors and markets,” said Whit McGraw, head of platforms at S&P Global Market Intelligence.
The move matters because it puts a ratings agency in the position of AI infrastructure provider, not just data seller, at the same time compliance and risk vendors are raising capital to build similar retrieval layers. The distinction that will decide who wins that fight is provenance: a deterministic query mode that can point to the exact rating record behind an AI-generated answer is a very different liability profile than a model that infers a credit view from training data, which is the failure mode regulators have flagged repeatedly around AI use in regulated risk decisions. Institutions adopting this kind of tool will need to ask which mode they are actually using before they rely on the output.
Source: PR Newswire
